1. Field of the Invention
This invention relates to a demodulating device, demodulating method and transmission medium, and in particular to a demodulating device, demodulating method and transmission medium which are suitable for demodulating a modulation code obtained by modulating data for application to data transmission or recording on a recording medium so as to reproduce data.
2. Description of the Related Art
When data is transmitted on a predetermined transmission path or recorded for example on recording media such as magnetic disks, optical disks, and magneto-optical disks, data modulation is performed which is suitable for the transmission or recording. One such type of modulation is known as block coding. This block coding converts a data sequence to blocks of m.times.i bit units (referred to hereafter as data words) , and this data word is converted to a code word comprising n.times.i bits according to a suitable code rule.
When i=1, this code is a fixed length code expressed by (d,k;m,n;1). When plural i are selected, a predetermined i is selected from the range 1 to imax (maximum value of i) and the conversion is performed, the code is a variable-length code. This block encoded code is represented by a variable length code (d,k;m,n;r).
Here, i is known as a restriction length, and imax is r (the maximum restriction length) . The minimum run d shows the minimum number of consecutive "0"s in repeated "1"s in the code sequence. The maximum run k shows the maximum number of consecutive "0"s in repeated "1"s in a code sequence.
In compact disks or mini-discs (trademark) etc., NRZI (Non Return to Zero Inverted) modulation, wherein "1" means inversion and "0" means non-inversion, is performed on the variable length code obtained as above, and the NRZI modulated variable length code (hereafter, referred to as a recorded waveform sequence) is recorded. This recorded waveform sequence will be referred to also as a level code.
When this level code is inverted so that "1" is replaced by "0" or "0" is replaced by "1", i.e. when reverse NRZI modulation is performed wherein "1" indicates an edge, the same code sequence as the original EFM code or RLL(1-7) code can be obtained. This reverse NRZI code sequence will also be referred to as an edge code.
Various modulation techniques have been proposed. If the minimum inversion interval of the recorded waveform sequence is Tmin and the maximum inversion interval is Tmax, to record at a high density in a linear velocity direction, the minimum inversion interval Tmin should be long, that is, the minimum run d should be large. From the clock reproduction aspect, moreover, the maximum inversion interval Tmax should be short, that is, the maximum run k should be small.
For example, one modulation technique used by magnetic disks or magneto-optical disks, etc., isRLL(2-7) The parameters of this modulation method are (2,7;1,2;3) If the bit interval of the recording waveform sequence is T, the minimum inversion interval Tmin (=(d+1)T) is 3(=2+1)T. If the bit interval of the data sequence is Tdata, this minimum inversion interval Tmin is 1.5(=(m/n)xTmin =(1/2).times.3)Tdata. The maximum inversion interval Tmax (=(k+1)T) is 8(=7 +1)T (=((m/n).times.Tmax)Tdata=(1/2).times.8Tdata =4.0Tdata). A detection window width Tw (=(m/n)T)) is 0.5(=1/2)Tdata.
Another modulation technique used by magnetic disks or magneto-optical disks, etc., is RLL(1-7). The parameters of this modulation method are (1,7;2,3;2). The minimum inversion interval Tmin is 2(=1+1) T (=2/3.times.2Tdata=1.33Tdata). The maximum inversion interval Tmax is 8(=7+1) T(=(2/3).times.8Tdata=5.33Tdata). Further, the detection window width Tw(=(m/n)XT) is 0.67(=2/3 data.
Comparing RLL(2-7) and RLL(1-7), for example in a magnetic disk system or magneto-optic disk system, to increase recording density in the linear velocity direction, RLL (2-7) for which the minimum inversion interval Tmin is 1.5 data preferable to RLL(1-7) for which the minimum inversion interval Tmin is 1.33 data. In practice, however, RLL(2-7) which has a larger detection window width Tw than RLL(2-7) and which is said to have a large tolerance to jitter, widely used.
The conversion table for the RLL(1-7) code is, for example, as follows.
TABLE 1 RLL (1, 7; 2, 3; 2) Data Code i = 1 11 00x 10 010 01 10x i = 2 0011 000 00x 0010 000 010 0001 100 00x 0000 100 010
Herein, the symbol x in the conversion table is given the value 1 when the following channel bit is 0, and given the value 0 when the following channel bit is 1 (same hereafter). The restriction length r is 2.
The conversion table for the RLL(2-7) code for which the minimum run d=2 and the maximum inversion interval Tmax is 8T (maximum run 7), is for example as shown below.
TABLE 2 RLL (2, 7; 1, 2; 3) Data Code i = 1 11 1000 10 0100 i = 2 011 001000 010 100100 000 000100 i = 3 0011 00001000 0010 00100100 The restriction length r is 3.
However, in a channel bit sequence which is modulated by RLL(1-7), the occurrence frequency of 2T which is Tmin is the greatest, followed by 3T and 4T. In general, if many periods occur wherein a large amount of edge information occurs early as in the case of 2T and 3T, this is advantageous for clock reproduction. However, if 2T occurs repeatedly, distortion of the recorded waveform occurs more easily. Specifically, the output waveform amplitude of 2T is small, and easily affected by defocusing or tangential tilt. Further, at a high linear density, recordings wherein the minimum mark is repeated are easily affected by external disturbances such as noise so that data reproduction errors tend to occur.
RLL(1-7) is often combined with PRML (Partial Response Maximum Likelihood), to improve S/N during playback of a high density recording. This method comprises, for example, Viterbi decoding equalized by PR(1,1) or PR(1,2,1) by matching the RF reproduction waveform to media characteristics. For example, a desirable reproduction output when equalization is performed by PR(1,1) is as follows.
 1 0 1 0 0 1 0 (channel bit data sequence) 1 1 0 0 1 0 0 (after NRZI conversion) . . . . . . 1 1 1 1 -1 -1 -1 -1 1 1 1 1 -1 -1 . . . . . . . . . +2 0 -2 0 +2 0 . . . (reproduction output)
The data after this NRZI conversion is level data. When the channel bit data is 1, it is given a different value (0 or 1) from the immediately preceding value (1 or 0) , and when the channel bit data is 0, it is given the same value (0 or 1) as the immediately preceding value (0 or 1). In this example, when the value after NRZI conversion is 1, "11" is decoded, and when the value after NRZI conversion is "0", "-1-1" is decoded. Waveform equalization when 2T which is Tmin is repeated, is performed to obtain this reproduction output. In general, waveform interference becomes longer the higher the linear density, therefore, waveform equalization also becomes longer as in PR(1,2,2,1) or PR(1,1,1,1).
However when the minimum run d=1 and a suitable waveform equalization is PR (1,1,1,1) as a result of high linear density, considering a situation when 2T which is Tmin occurs repeatedly, the reproduced signal at that time is
 1 0 1 0 1 0 1 0 1 0 (channel bit data sequence) 1 1 0 0 1 1 0 0 1 1 (after NRZI conversion) . . . . . . 1 1 1 1 1 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1 1 1 1 1 1 1 1 1 -1 -1 -1 -1 . . . . . . . . . 0 0 0 0 . . . (reproduction output).
and the zero level will be traced for a long time.
This shows that a situation when practically no signal level is output after waveform equalization continues, and therefore Viterbi decoding does not merge. This also causes considerable loss of data reproduction or clock reproduction stability.
This kind of channel bit data sequence, for example in the case of RLL (1,7;2,3;2) in Table 1, occurs when the premodulated data sequence is "10-01-10-01-10- . . . "
Similarly, in the case of RLL (2,7;1,2;3) in Table 2, the premodulated data sequence is "010-010-010-010- . . . "
In this regard, the inventors already proposed the use of a code for limiting repetition of Tmin in Patent Application No. Hei 9-133379.
According to this proposal, when the variable length code (d,k;m,n;r) is, for example, a variable length code (1,7;2,3;3), i.e. when d which is the minimum run of "0" is 1 bit, k which is the maximum run of "0" is 7 bits, m which is basic data length is 2 bits, n which is the basic code length is 3 bits, and r which is the maximum restriction length is 3, the conversion table is such as is shown for example in the following Table 3.
TABLE 3 RML (1, 7; 2, 3; 3) Data Code i = 1 11 00x 10 010 01 10x i = 2 0011 000 00x 0010 000 010 0001 100 00x 0000 100 010 i = 3 100110 100 000 010 The restriction length r is 3.
In the above Table 3, when the data sequence is "10", and in particular when the following four bits of data are looked up so that the total six (bit) data sequence is "100110", by converting data to a code which limits repetition of the minimum run, the minimum run can be repeated up to 5 times by the modulation of Table 3.
Comparing with RLL(1-7) of Table 1, the minimum run and maximum run are the same, and the conversion ratio m/n (ratio of data words and converted code words) is the same, but the restriction length has increased from 2 to 3. This shows that the maximum size of the table has increased, and shows that error propagation often increases when, for example, a bit shift error occurs during demodulation.
A bit shift error is an error wherein a "1" representing the edge in the code sequence is shifted one bit forward or backward. The error propagation is expressed as a number of bits from the start to the end of a demodulation error produced when a code sequence, in which an error occurs at one position due for example to a bit shift error, is decoded without modification.
This bit shift error is the form of error which occurs most frequently during data reproduction in an actual recording/playback device, and it has been found to occur exclusively in the vicinity of the minimum run.
As an example of modulation of codes other than these, when the variable length code (d,k;m,n;r) is a variable length code (2,7;1,2;4), i.e. when d which is the minimum run of "0"is 2 bits, k which is the maximum run of "0" is 7 bits, m which is the basic data length is 1 bit, n which code length is 2 bits, and r which is the maximum restriction length is 4, the conversion table is such as is shown for example in the following table.
TABLE 4 RML (2, 7; 1, 2; 4) Data Code i = 1 11 1000 10 0100 i = 2 011 001000 010 100100 000 000100 i = 3 0011 00001000 0010 00100100 The restriction length r is 4.
In the above Table 4, when the data sequence is "010", and in particular when the following two bits of data are looked up so that the total five (bit) data sequence is "01001", by converting data to a code which limits repetition of the minimum run, the minimum run can be up to 4 times by the modulation of Table 4.
Comparing with RLL(2-7) of Table 2, the minimum run and maximum run are the same, and the conversion ratio m/n is the same, but the restriction length has increased from 3 to 4. This shows that the maximum size of the table has increased as described above, and shows that error propagation often increases when, for example, a bit shift error occurs during demodulation.
When, as described above, recording media such as magnetic disks, magneto-optic disks or optical disks are recorded at high density, and codes with a long minimum run such as RLL(1-7) or RLL(2-7) are selected as modulation codes, if the minimum inversion interval Tmin occurs repeatedly, recording and playback distortion occurs easily which is disadvantageous for clock reproduction.
When high linear density (recording) is performed and PR(1,1,1,1) equalization is performed with a d=1 code, if the minimum inversion interval Tmin is repeated, the logic of the reproduced signal outputs zero continuously and Viterbi decoding does not merge, which is disadvantageous for clock reproduction.
This invention, which was conceived in view of the above problems, therefore adds a code which limits the minimum inversion interval Tmin from repeating for a long time, to a conversion table as in the prior art, e.g. RLL(1-7) or RLL(2-7), and performs demodulation processing by a table RML(1-7) method or RML(2-7) method whereby clock reproduction is rendered more stable.
However, this RML(1-7) method has a longer restriction length than RLL(1-7) of the prior art and RML(2-7) has a longer restriction length than RLL(2-7) of the prior art, so when bit shift errors occur in the reproduced data obtained from a recording/playback device, error propagation during data decoding becomes longer.